In most practical applications, it's common to utilize multiple features from different views to represent one object. Among these works, multi-view subspace-based clustering has gained extensive attention from many researchers, which aims to provide clustering solutions to multi-view data. However, most existing methods fail to take full use of the locality geometric structure and similarity relationship among samples under the multi-view scenario. To solve these issues, we propose a novel multi-view learning method with locality relationship constraint to explore the problem of multi-view clustering, called Locality Relationship Constrained Multi-view Clustering Framework (LRC-MCF). LRC-MCF aims to explore the diversity, geometric, consensus and complementary information among different views, by capturing the locality relationship information and the common similarity relationships among multiple views. Moreover, LRC-MCF takes sufficient consideration to weights of different views in finding the common-view locality structure and straightforwardly produce the final clusters. To effectually reduce the redundancy of the learned representations, the low-rank constraint on the common similarity matrix is considered additionally. To solve the minimization problem of LRC-MCF, an Alternating Direction Minimization (ADM) method is provided to iteratively calculate all variables LRC-MCF. Extensive experimental results on seven benchmark multi-view datasets validate the effectiveness of the LRC-MCF method.